Here a few pointers which may be able to help you:
- For every pixel position in the output, the algorithm tries to find the corresponding position in the input image. For subpixel shifts, such a pixel position may not exist. In such cases we use interpolation techniques to find the intensity value. In the case of ‘imtranslate’ there are techniques available: ‘cubic’, ‘linear’ and ‘nearest’.
- For ‘cubic’and ‘linear’, the intensity value is calculated as a function of the neighboring values. In this case it may not considered as a rounding-off process.
- But for ‘nearest’ method, the translation may be considered as being approximated to the nearest full pixel increment.
Hope this helps!!